Patent Application
20230123838
The present invention relates generally to electrophysiological (EP) signal analysis, and particularly to the analysis of cardiac arrhythmias.
Clustering refers to the task of identifying groups or clusters in a data set. In density-based clustering, a cluster is a set of data objects spread in the data space over a contiguous region of a high density of objects. Density-based clusters are separated from each other by contiguous regions of a low density of objects. Data objects located in low-density regions are typically considered noise or outliers.
Methods to characterize the propagation of activation signals in cardiac tissue were previously proposed in the patent literature. For example, U.S. Pat. No. 9,737,227 describes a system and method for mapping a cardiac anatomical structure that includes sensing activation signals of physiological activity with a plurality of mapping electrodes disposed in or near the anatomical structure. Patterns among the sensed activation signals are identified based on a similarity measure generated between each unique pair of identified patterns which are classified into groups based on a correlation between the corresponding pairs of similarity measures. A characteristic representation is determined for each group of similarity measures and displayed as a summary plot of the characteristic representations.
As another example, U.S. Patent Application Publication No. 2010/0280400 describes a cardiac rhythm management system that can be used to detect episode beats associated with cardiac events in a subject's body. These events may be monitored and depolarization morphology information can be derived for candidate arrhythmic beats in an arrhythmia episode. An arrhythmic beat morphology template may be formed from selecting at least one of the candidate arrhythmic beats based upon user's labeling according to specific morphologies of one or more candidate episodes. Methods of use are also presented.
U.S. Patent Application Publication No. 2004/0176697 describes an analysis of electrocardiograms (ECGs) during atrial fibrillation. In particular, the invention relates to the use of such methods for the creation and validation of cardiac models and use in the refined diagnosis of heart disease. Classification of the atrial fibrillation of a given patient or group of patients is performed using an auto regressive (AR) model with coefficients subjected to a mathematical cluster analysis, including hierarchical methods (e.g., single linkage, average linkage (weighted and unweighted), centroid, median and complete linkage) and non-hierarchical methods, e.g., the k-means clustering algorithm, adaptive k-means, k-medoids, and fuzzy clustering.
An embodiment of the present invention that is described hereinafter provides a method for clustering of electrophysiological signals using similarities among arrhythmogenic activations, including defining multiple different types of arrhythmias. Similarity measures are defined for the types of arrhythmias. A set of EP signals is received, the set acquired in a heart of a patient. Using the similarity measures, the set of EP signals is partitioned into at least two clusters, each cluster containing EP signals complying with a respective similarity measure for a respective type of arrhythmia. The clusters are indicated to a user.
In some embodiments, indicating the clusters includes presenting a representative EP signal of each cluster.
In some embodiments, partitioning the set of EP signals includes assigning the EP signals to the clusters based on a number, N, of activations in each EP signal over a given window of interest.
In an embodiment, partitioning the set of EP signals includes assigning the EP signals to the clusters based on a number of cross iso-peaks within an activation in each EP signal.
In another embodiment, partitioning the set of EP signals includes assigning the EP signals to the clusters based on a width of an activation in each EP signal.
In some embodiments, partitioning the set of EP signals includes assigning the EP signals to the clusters based on a peak-to-peak bi-polar voltage of an activation in each EP signal.
In some embodiments, partitioning the set of EP signals includes assigning the EP signals to the clusters based on a sharpest slope of an activation in each EP signal.
In an embodiment, the similarity measure is one of a linear sum function and a quadratic sum function.
In some embodiments, the EP signals are electrograms.
There is additionally provided, in accordance with another embodiment of the present invention, a system for clustering of electrophysiological signals using similarities among arrhythmogenic activations, including a memory and a processor. The memory is configured to store (i) definitions multiple different types of arrhythmias, and (ii) definitions of similarity measures for the types of arrhythmias. The processor is configured to (a) receive a set of EP signals acquired in a heart of a patient, (b) using the similarity measures, partition the set of EP signals into at least two clusters, each cluster containing EP signals complying with a respective similarity measure for a respective type of arrhythmia, and (c) indicate the clusters to a user.
The present invention will be more fully understood from the following detailed description of the embodiments thereof, taken together with the drawings in which:
The electrical activity of tissue of an organ of a patient, such as a cardiac chamber, can be mapped, e.g., electroanatomically and/or electrophysiologically, using a mapping catheter having suitable electrodes fitted at its distal end. An EP mapping system may acquire and analyze the signals output by the catheter.
In practice, during a cardiac arrhythmia analysis procedure, a very large number, for example, up to few tens of thousands or even more EP signals (e.g., bi-polar electrograms, may be acquired. Within these numerous EP signals, a physician may be interested in looking only at a small subset of the EP signals, for example electrograms showing some special fractionation, or electrograms characterized by a particular cycle length or local activation time (LAT). However, because of the large number of acquired EP signals, it is both difficult and time consuming to locate members of a subset among the entire set of acquired EP signals.
Embodiments of the present invention that are described hereinafter provide algorithms that automatically divide a set of EP signals into clusters (EP signals may be electrograms (EGM) or electrocardiograms (ECGs), for example). In an exemplary embodiment, a set of electrograms is acquired from a patient with arrhythmia. The electrograms are then analyzed to find parameters such as activation times, activation duration, number of peaks in an electrogram, slopes of the peaks, and the like. An algorithm uses these parameters to associate each electrogram with one of multiple types of predefined EP activations, such as normal conduction, atrial fibrillations, flutter, tachycardia, or other type of arrhythmia.
For each type of EP signal, a “similarity measure” is predefined, being the distance relating EP signals (e.g., electrograms) within the defined type of EP signal (e.g., type of arrhythmia). This mathematical similarity measure, for example one defined by a similarity metric as described below, delineates the bounds of the parameters defining “neighboring” signals. As an example, electrograms for a particular type of arrhythmia may be defined as those having two activations, each peak having a maximum voltage within some specified range, e.g., between 2 mV and 5 mV. An electrogram is deemed in the neighborhood of a given electrogram of this type if its two activations are within a given similarity difference (e.g., 10%) of those of the given electrogram.
In some embodiments, a processor runs a clustering algorithm that uses the predefined types and their respective distances to decide if a particular EP signal is clustered within one of the defined types. Typical clustering algorithms that may be used include the DBSCAN and the OPTICS algorithms, which were previously described in the literature. The two algorithms are described, for example, in a review paper entitled “Density-based clustering,” by Kriegel et al., WIREs Data Mining and Knowledge Discovery, Volume 1, Issue 3, May/June 2011, pages 231-240 of, which is incorporated herein by reference.
In some embodiments, the processor clusters the EP signals automatically without user intervention. Subdividing the EP signals into clustered groups (i.e., types) of signals, for example ten groups, reduces the amount of time a physician may need to examine the EP signals. Thus, the physician may only need to inspect ten cluster-type representative signals, rather than many or all of the acquired signals.
Catheter 29 may be further used to perform an ablation, such as a radiofrequency (RF) or irreversible electroporation (IRE).
The respective locations of mapping-electrodes 22 inside heart 23 (i.e., where the electrograms are measured) are also tracked, such that a processor 28 may link each acquired electrogram with the location at which the signal was acquired. System 21 externally senses electrical position signals using a plurality of external electrodes 24 coupled to the body of patient 25; for example, three external electrodes 24 may be coupled to the patient's chest, and another three external electrodes may be coupled to the patient's back. For ease of illustration, only one external electrode is shown in
An example of a system capable of using the sensed electrical position signals to track the locations of mapping-electrodes 22 inside heart 23 of the patient is the CARTO® 3 system (produced by Biosense Webster, Irvine, Calif.). The CARTO® 3 system uses a tracking method known as Advanced Current Location (ACL), which is described in detail in U.S. Pat. No. 8,456,182 whose disclosure is incorporated herein by reference.
A data acquisition module 33 receives the multiple electrograms conveyed to an electrical interface 35 over a wire link that runs in catheter 29. Processor 28 of system 21 receives these cardiac signals via the electrical interface 35, and uses the disclosed clustering algorithm based on the aforementioned similarity function to cluster these signals according to different predefined arrhythmia types, e.g., atrial fibrillations, tachycardias, flutter and more, which may amount, as in the example shown in
The definitions of the various types of arrhythmia, and the corresponding definitions of the similarity measures, are typically stored in a memory 31 for use by processor 28. Processor 28 also stores the clustered electrograms in memory 31 for further analysis, such as for constructing an EP map.
The exemplary embodiment shown in
Processor 28 typically comprises a general-purpose computer with software programmed to carry out the functions described herein. The software may be downloaded to the computer in electronic form, over a network, for example, or it may, alternatively or additionally, be provided and/or stored on non-transitory tangible media, such as magnetic, optical, or electronic memory. In particular, processor 28 runs a dedicated algorithm that enables processor 28 to perform the steps described in
As noted above, a clustering algorithm that is provided analyzes acquired EP signals to find parameters of the EP signals such as activation times, activation duration, number of peaks in an EP signal, and slopes of the peaks. The processor utilizes an algorithm that uses these parameters in a similarity measure to associate each EP signal, if found aberrative, with different predefined arrhythmia types.
In an exemplary embodiment, by-type clustering parameters for atrial flutter arrhythmia (for the similarity function) are:
Number of activations within the window of interest (WOI): N
Number of cross iso-peaks within an activation: P
Width of an activation: W
Peak to peak bipolar voltage of an activation: V
Sharpest slope of an activation: S
Between a pair of EP signals, a normalized value of the similarity function within the range of [0,1] is correspondingly calculated using the differences: dP, dW, dV, dS.
Assuming a stage clustering model, a first stage is characterized solely by a number N of activations within the WOI, and produce four clusters (flat: 0 activations; single potential, double potential, other: more than two activations).
The second stage of the model is calculated based on a similarity function F(EP_signal_1, EP_signal_2) on the rest of the parameters:
F(EP_sig_1,EP_sig_2)=0.5·dP+0.25·dS+0.125·dV+0.125·dW
The above linear similarity function is brought by way of example. Other functions, or norms, for example, quadratic ones in dP, dW, dV, and dS, may be used.
In set 41, different graphs may have a different number of activations 42, N, within a common WOI 202. For example, EP signals types 2 and 4 have N=2 activations 42 (e.g., peaks, complexes), whereas EP signals types 6 and 9 have N=3 peaks. In that regard, EP signal type 3 is unique by having N>4 peaks.
As another example, EP signals type 7 shows a large width of an activation, W, whereas EP signals types 4 and 10 show a narrower W.
A physician may need to inspect only set 41 of the ten-cluster representative EP signals types 1 to 10 in order to assess the cardiac activity and to determine if and which of the types has clinical significance. For example, most types may represent normal cardiac activities with only a small number of types representing suspicious electrical activity, abnormal substrate, possible ablation target, or other morphology of interest to a physician.
The graphs of
Next, the algorithm selects a similarity measure defined for each arrhythmia type, with physician 27 optionally adjusting parameters of the selected one or more similarity measures, at similarity measures defining step 304.
At an EP data receiving step 306, processor 28 receives a set of thousands of EP signals acquired by a multi-electrode catheter (e.g., catheter 29) that mapped (or is mapping in real-time) a cardiac chamber.
Using the one or more similarity measures, at a clustering step 308, processor 28 partitions the set of EP signals into at least two clusters, each cluster characterized by its EP signals complying with the respective similarity measure for the type of arrhythmia.
Finally, processor 28 presents on display 26 representative graphs, such as set 41, of the different clusters, at EP signals presentation step 310. Typically, the processor selects a median signal from each cluster of signals as a representative signal for the cluster. Alternatively, however, any other suitable selection can be used.
Although the embodiments described herein mainly address cardiac applications, the methods and systems described herein can also be used in other applications, such as in mapping of electrical activity in the brain.
It will thus be appreciated that the embodiments described above are cited by way of example, and that the present invention is not limited to what has been particularly shown and described hereinabove. Rather, the scope of the present invention includes both combinations and sub-combinations of the various features described hereinabove, as well as variations and modifications thereof which would occur to persons skilled in the art upon reading the foregoing description and which are not disclosed in the prior art. Documents incorporated by reference in the present patent application are to be considered an integral part of the application except that to the extent any terms are defined in these incorporated documents in a manner that conflicts with the definitions made explicitly or implicitly in the present specification, only the definitions in the present specification should be considered.
